Do Patients Regret Having Received Systemic Treatment for Advanced Non‐Small Cell Lung Cancer: A Prospective Evaluation

Abstract

Background

Thousands of patients annually receive treatment for advanced non‐small cell lung cancer (NSCLC), but little is known about their views on the decision to receive that treatment, or regret. This trial prospectively evaluated the incidence of regret and whether baseline characteristics, patient decision‐making parameters, or clinical progress early in the treatment course predicts regret.

Materials and Methods

Patients receiving systemic treatment for advanced NSCLC completed every 3‐week patient reported outcome (PRO) assessment using the electronic Lung Cancer Symptom Scale (eLCSS‐QL), including the 3‐Item Global Index (3‐IGI; assessing overall distress, activities, and quality of life [QL]). A prespecified secondary aim was to determine the frequency of regret evaluated at 3 months after starting treatment. Patients were randomized to usual care or enhanced care (which included use of the DecisionKEYS decision aid).

Results

Of 164 patients entered, 160 received treatment and 142 were evaluable for regret. In total, 11.5% of patients and 9% of their supporters expressed regret. Baseline characteristics did not predict regret; regret was rarely expressed by those who had a less than 20% decline or improvement in the 3‐IGI PRO score after two treatment cycles. In contrast, when asked if they would make the same decision again, only 1% not having a 20% 3‐IGI decline expressed regret, versus 14% with a 3‐IGI decline (p = .01).

Conclusion

The majority of patients having regret were identified early using the PRO 3‐IGI of the eLCSS‐QL measure. Identifying patients at risk for regret allows for interventions, including frank discussions of progress and goals early in the treatment course, which could address regret in patients and their supporters.

Implications for Practice

This report documents prospectively, for the first time, the incidence of treatment‐related regret in patients with advanced lung cancer and outlines that risk of regret is associated with patient‐determined worsening health status early in the course of treatment. Identifying patients at risk for regret early in treatment (before the third cycle of treatment) appears to be crucial. Counseling at that time should include a discussion of consideration of treatment change and the reason for this change.

Short abstract

Considering that most patients are unlikely to have a major response treatment for advanced non‐small cell lung cancer, efforts should be made to avoid or reduce patient regret associated with treatment decision making. This article focuses on frequency of regret, prospectively evaluated in patients at 3 months after starting systemic treatment for advanced lung cancer.

Introduction

Although major advancements have occurred in the treatment of advanced lung cancer, patients continue to be faced with difficult realities, challenges, and decisions. With the rapid pace of unresectable lung cancer and with median overall survival times around the 12‐month mark in studies including patients with good performance status (PS), treatment decisions often need to be made rapidly. These factors make quality decision making especially difficult. Regret may result from poor decision making, and most importantly, negative treatment outcomes may also lead to regret with the treatment choice. Given that the majority of patients are unlikely to have a major response in advanced non‐small cell lung cancer, consideration should be made to avoid or reduce regret.

In a 2017 New England Journal of Medicine article focused on regret, the authors emphasized that the risk of regret exists in almost every medical decision a patient makes [1]. Additionally, the risk of regret is underestimated in patient decision making, and the impact of regret can be especially high in making decisions about major illnesses and their treatment. The authors conclude that anticipation of the potential for regret should be considered much more by physicians recommending treatment and helping patients make decisions. Some prospective trials have indicated that early involvement of supportive and palliative care specialists can improve quality of life and even affect survival [2]. Many factors may have contributed to these benefits, including better symptom control and possibly help in better decision making concerning continuing treatment. The concept of informed, shared decision making promotes meaningful dialogue between physician and patient as opposed to “a unidirectional, dutiful disclosure” of alternatives, risks, and benefits” [3].

Prior investigations have outlined patient uncertainties about prognosis. A qualitative study of 35 patients with advanced lung cancer indicated changing degrees of optimism about recovery during the course of treatment [4]. It was reported that physicians generally did not follow through with clarity of the results of ongoing treatment, especially when the patient or caregiver did not pursue more information about effectiveness of the treatment. Regret, which may result from this lack of clarity, may add not only to the patient's burden but also to that of the family and especially that of the patient's close supporters.

Surprisingly, in a malignancy as common as lung cancer, the incidence of regret and associated factors has not been determined. Recognizing that there may be consequences of poor decision making, a 2019 review of the topic of shared decision making found 35 studies in oncology. Although five of these emphasized lung cancer, none evaluated regret [5]. In contrast, a 2016 systematic review of 59 studies dealing specifically with regret failed to find a single one focused on lung cancer [6].

In this current study, the frequency of regret was prospectively evaluated at 3 months after starting systemic treatment for advanced lung cancer. The intent was to analyze whether (a) baseline characteristics, including quality of life and patient‐reported outcomes, (b) patient decision‐making parameters, or (c) clinical progress as expressed by the patient very early in the treatment course could predict who would ultimately experience regret during treatment.

Materials and Methods

Design and Setting

All patients were enrolled in a prospective randomized controlled trial (RCT). Patients were receiving first‐ or subsequent‐line systemic therapy (chemotherapy or a checkpoint inhibitor) for stage IIIB or IV non‐small cell lung cancer (NSCLC). Treatment choice was at the discretion of the treating physician. Patients receiving molecularly targeted tyrosine kinase inhibitor agents were not eligible. The delivery of the intervention at three cancer centers included 16 physicians and 4 research nurses as investigators at five hospitals. The primary aim of the RCT was to determine if immediate availability of changes in patient reported outcomes (PROs) would affect the amount of chemotherapy and imaging received [7]. A prespecified secondary aim, the subject of this paper, was to determine the frequency of regret evaluated at 3 months after starting treatment for all study patients and to find if this could be identified early (after just the second treatment cycle).

Study Sample

Patients were included in the analysis if they met the modified intention‐to‐treat (MITT) criteria: informed consent, randomized, evaluable baseline data, and completion of at least one treatment cycle. All patients completed a quality of life (QL)–PRO assessment at baseline and every 3 weeks. Patient supporters were also asked to participate in the study. All consenting supporters were asked to complete this same PRO measure from their perception of the patient's status. Patients and supporters completed a decision conflict scale at baseline and a decision regret scale at 10–12 weeks after starting treatment.

Study Intervention

The intervention included a decision aid coupled with immediately available QL‐PRO results over the time of the study. Patients were randomized to the intervention arm (enhanced care arm) or to usual care, on a 1:1 basis. All patients completed the electronic Lung Cancer Symptom Scale (eLCSS‐QL) every 3 weeks, but only in the intervention, enhanced care, arm were the results given to the treating physician.

Patients randomized to the enhanced care arm also completed a decision aid (“DecisionKEYS for Balancing Choices”), which uses a four‐cell decision balance sheet of positives and negatives related to values for the patient and for others [8, 9]. Results of the decision aid were immediately shared with the treating team.

Primary Measures

Decision regret scale (DRS). This 5‐item scale uses a 5‐point Likert scoring method and has published psychometrics [10]. Regret was defined as the selection of the top two negative response categories (“disagree” and “strongly disagree”), indicating the greatest extent of regret for any of the 5 questions.

Decisional conflict scale (DCS). This 16‐item scale uses a 5‐point Likert score for measuring difficulty in making decisions [11]. The traditional 16‐item form was used based on published support for validity. For the uncertainty subscale, the score was converted to the equivalent of a 1–100 scale by dividing the score by the number of items summed, subtracting 1, and then multiplying by 25. The higher the score, the more decisional conflict there was.

The eLCSS‐QL. All patients were administered a quality of life instrument with published psychometric properties [12]. The LCSS has a total of nine items; six are common symptoms of lung cancer, and three are the global items included in the 3‐Item Global Index (3‐IGI). The items included in the 3‐IGI are overall distress, interference with normal activities, and overall health‐related quality of life (Fig. 1). Each item is determined by a visual analog scale: the 3‐IGI is the sum of these three summary scores. More than 80% of patients have three or more of these six symptoms, and all patients have useful scores in each of the 3 global items [13]. The well‐tested paper version of the LCSS has been programmed for use with a computer‐assisted portable device (handheld pocket PC or tablet) in an electronic version (eLCSS‐QL) to enhance ease of use and to overcome barriers in incorporating evaluation of PROs as part of clinical management [12]. The electronic version has shown high correlation with the paper version. This PRO assessment takes about 2 minutes, has high patient, nurse, and physician acceptance, and allows for immediate graphic demonstration of the results over time [12].

Figure 1.

Lung Cancer Symptom Scale patient‐reported outcome measure: a total of nine items including the Average Symptom Burden Index and the 3‐Item Global index. Analyzed using 100‐mm visual analogue scales [21, 22].

Study Procedures

Approval was obtained from Human Investigations Committees at each site. Written informed consent was obtained from each participant (and supporter) by the physician or the research nurse, who also administered the battery of measures.

For this analysis related to regret, patients received or were monitored for three outcomes: (a) for the PRO assessment, the electronic Lung Cancer Symptom Scale (eLCSS‐QL) was used to assess the patient's quality of life and symptoms at baseline and every 3 weeks for all patients; (b) the decision aid was administered at baseline to the patients randomized to the enhanced care arm; (c) the Decision Conflict Scale was administered at baseline to all patients; and (d) at 10–12 weeks after starting treatment, all patients completed the Decision Regret Scale. All PRO assessments, including the eLCSS‐QL, the DecisionKEYS balance sheet, the decision conflict scale, and the decision regret scale, were completed prior to patient visits with the physician, and prior to receiving results of radiographic studies or other tests.

For the intervention group (enhanced care arm), the interactive decision aid was used with a decisional balance sheet for each treatment decision. This decision aid is based on the Janis and Mann conflict theory of decision making [14, 15]. This theory, used as a brief tutorial diagram for the intervention group, predicts decision‐making behavior for consequential decisions, those which are emotionally laden and motivationally driven, which then results in satisfaction or regret. The physician reviewed the results of the LCSS symptomatic and quality of life scores over time using a graph at each visit. The physician and the nurse then discussed results of the decision balance sheet with the patient or patient–supporter pair.

Statistical Analysis

A stratified block randomization design was used in which strata were determined by two variables for the two‐group design: line of therapy and presenting quality of life score.

Descriptive statistics were used to describe the incidence, extent, and associated factors concerning regret expressed at 10–12 weeks. χ² tests and Fisher's exact tests were used to assess the difference in proportions between the response variable, regret at 10–12 weeks, and treatment group (enhanced care [EC]; usual care [UC]), the percent change in 3‐IGI, and in other categorical explanatory variables (e.g., decisional conflict). Wilcoxon rank sum tests also were used to test for difference in distributions of continuous explanatory variables for 10–12 week decisional regret. The Kaplan‐Meier product limit method was used to estimate survival and the log‐rank test was used to compare survival curves between groups of interest. Results were considered significant at a 5% significance level. Analyses were conducted in SAS v9.4 (SAS Institute, Inc. Cary, NC).

Results

Demographics and Health Status

Of the 164 patients with advanced NSCLC who signed consent to participate in the study, 160 received treatment. Thus, for this RCT, 160 of the 164 (98%) of the randomized patients met the MITT criteria for adequacy for evaluation. At the 10–12‐week mark after randomization, 12 patients had died, and 5 were too ill to be able to fill out the decision regret scale. These occurrences reduced the evaluable sample to a potential of 143 for the regret endpoint at 10–12 weeks (specifically, DRS completion time interquartile range: 25th percentile, 9 weeks and 75th percentile, 11.6 weeks). In addition, 13 patients declined to complete the DRS (9%), resulting in a total of 130 patients (91% of patients possible to evaluate at 10–12 weeks, and 81% of the MITT population).

Baseline characteristics, with clinical, social, decision making, and PRO factors, are seen in Table 1. Patients are listed by those who expressed regret and those who did not at 10–12 weeks following randomization. As seen in the table, most baseline factors were fairly similar between those who did or did not express regret at 10–12 weeks, with no significant differences found.

Table 1.

Baseline characteristics of patients expressing regret at 10–12 weeks: Clinical and other factors

Factors	Patients with regret (n = 15; 11.5%)	Patients without regret (n = 115; 88.5%)
Clinical Factors, %
Age ≥70	33%	27%
Male	67%	54%
Stage IV	100%	93%
First‐Line Treatment	80%	74%
ECOG PS = 1	60%	63%
Social Factors, %
Not married	47%	57%
No designated supporter	53%	38%
Decision‐making factor
Decision uncertainty score (median) (higher score = more conflict)	25 (n = 14) (IQR: 0 to 42)	17 (n = 111) (IQR: 0 to 25)
Patient‐Reported Outcome Factors, n (IQR)
Global distress score (0 = worst;100 = best)	76 (47–95)	76 (50–94)
Global activity level score (0 = worst; 100 = best)	66 (28–94)	61 (43–91)
Global quality of life score (0 = worst; 100 = best)	64 (47–96)	66 (46–91)
3‐IGI score (0 = worst; 300 = best)	198 (124–279)	196 (152–254)

Percentages are given for categorical factors; medians and interquartile ranges (IQR) are given for continuous factors. No significant differences were found in the baseline values between the two groups for any characteristic.

Patients having regret were more often not married, less likely to have a supporter, and had higher (poorer) decision uncertainty subscale scores. Of note, all PRO factors at baseline were very similar between the two groups, as was the percentage of patients with Eastern Cooperative Oncology Group (ECOG) PS = 1 (61% ± 1%), the most frequent performance status group in the study.

Regret and Early Change in eLCSS‐QL 3‐IGI

Of the 130 patients completing the DRS at 10–12 weeks after randomization, 15 (11.5%) expressed regret. No baseline characteristic predicted the likelihood of the patient having regret at 10–12 weeks (Table 1). Nonetheless, decline in PROs (specifically, the 3‐IGI) early in the treatment period (after just 2 treatments) was highly predictive of the patient having regret thereafter. Of the 130 patients completing the DRS, 125 of these patients also completed the 3‐IGI PRO in this time frame.

Prior reports in patients with NSCLC or with mesothelioma demonstrated that the 3‐IGI at baseline predicts eventual survival more accurately than ECOG PS. To that end, within ECOG PS = 1, significantly different survival was seen if the patient 3‐IGI scores are analyzed as low (lowest tertile), medium (midtertile) or high (highest tertile) [16]. Additionally, recent results indicated that patients with a 20% decline from baseline in the 3‐IGI after just two treatment cycles (typically cycles started on days 1 and 22, with evaluation 3‐weeks later, prior to the third planned treatment), have significantly poorer survival outcomes when compared with those with better 3‐IGI results [7]. Baseline 3‐IGI scores (overall distress, interference with normal activities, and health‐related quality of life) were very similar for those who ultimately expressed regret at 10–12 weeks and for those who did not (Table 1).

In contrast, although no baseline characteristic predicted regret, a decline in perceived health status as expressed by the patient early in the treatment course was significantly associated with a higher indication of regret at 10–12 weeks (about 3 months). Specifically, those patients with a 20% decline in the 3‐IGI after just 2 treatment cycles, when assessed at the 6‐week point after starting treatment, expressed regret four to ten times more often than those experiencing a lesser impact on the 3‐IGI (Table 2). This group of 29 patients with a 20% 3‐IGI decline at 6 weeks expressed regret far more often as determined by each of the five DRS questions (Table 2). The key question for this analysis using the DRS concerned the single item asking if the patient would make the same decision again. It is notable for this item that only 1% not having a 20% decline in the 3‐IGI after 2 treatment cycles ultimately expressed regret at 10–12 weeks, as contrasted to 14% of those with a greater than 20% 3‐IGI decline (p = .01). Results were similar in both randomization groups.

Table 2.

Results for regret (assessed at 10–12 weeks) predicted by change in the 3‐IGI score, measured at the 6‐week time point*

Decision Regret Scale Items	≥20% decline in 3‐IGI (n = 29; 22%), %	Improvement or <20% decline in 3‐IGI (n = 96, 78%), %	p value
I would not make same choice again	14	1	.01
I regret the choice that was made	17	3	.02
The choice did me a lot of harm	14	4	.08
The decision was not a wise one	10	1	.04
It was not the right decision	3	2	NS

As regret is a negative concept, all items are expressed here as negatives.

^*Change in 3‐IGI score at 6 weeks compared with baseline.

Survival and Regret at 10–12 Weeks

Those patients expressing regret 10–12 weeks after randomization had a much poorer survival rate at 1 year when compared with patients without regret (32% vs. 61%, p = .038). This result is consistent with the findings that the majority of these patients had at least a 20% decline in the 3‐IGI, as determined 6 weeks after randomization, and that this decline in the 3‐IGI had previously been associated with poorer survival when compared with patients with better 3‐IGI scores at that time [7, 16].

Regret among Patients' Supporters

In this trial, patients were requested to designate a chosen supporter if available (defined as “a family member or concerned other chosen by the patient”) as a participant. There were 69 such supporters designated. Of the 130 patients who completed the DRS, 43 supporters (33%) also completed the DRS. Of these 43 patient‐supporter pairs, there was 88% agreement with respect to having regret.

Decision Aid Intervention Group Differences Including Evaluation

Participants for this intervention RCT included 160 patients with advanced NSCLC, their 69 supporters, and their health professionals (a total of 16 physicians and 4 research nurses). For the DCS at baseline, the EC group mean score was 15.56 (95% confidence interval [CI], 12.89–18.23); the UC group mean score was 17.68 (95% CI, 15.07–20.29). Three results of note included the following: (a) no significant difference was found between decision conflict total score and intervention group at baseline (p = .25); (b) no association was found for the DCS total score over time, when the baseline pretest was compared with the approximately 3‐month post‐test score (p = .19); and (c) no significant differences were found between the EC and UC groups for decision regret. Only the EC group used the interactive decision aid. Additionally, as in Table 1, for the uncertainty subscale score alone, the baseline scores were only slightly higher for those with regret (median score of 25) than for those without regret (median score of 17).

Discussion

Regret, like depression and anxiety, is another emotional or serious mental health issue faced by many patients with cancer. With lung cancer, the burden of regret is even higher in that many patients already feel guilt or regret over a long tobacco usage history [17]. Although systemic treatment for lung cancer has a modest positive effect on survival and on symptoms for an overall treatment group, these benefits do not occur for all. Given the underlying regrets frequently found in patients with lung cancer, it is not surprising that the choice of treatment could ultimately result in regret as well. To reduce suffering in this difficult malignancy and to maximize patient specific care, attention to mental health aspects of the disease and treatment is mandatory. Treatment regret is not specific to lung cancer, but with the fact that lung cancer is the most common cause of cancer‐related death, attention to regret in this malignancy is reasonable and may be instructive for other cancers.

A central finding of this prospective study was the 11.5% incidence of treatment‐related regret as expressed by patients. No prior study, retrospective or prospective, had determined the occurrence rate of regret related to treatment choice. It was also interesting to note a similar rate of regret was reported by supporters of the patients. Although symptoms contribute to suffering of patients and their supporters, regret likewise affects both groups, and supporters may be left with dealing with regret after the patient has died.

It may be viewed positively that almost 90% of patients did not express regret at 3 months after starting systemic treatment, especially when considering that many treatment regimens are associated with challenging side effects. This finding of low rates of expressed regret is consistent with the 2016 systematic review of 59 studies dealing with regret, in which two‐thirds (66%) of the studies included were in cancer settings [6].

Focusing on those who experienced treatment‐related regret is a key approach to lessening or preventing regret among patients and their supporters. Baseline characteristics, including typical demographics plus social and quality of life findings, did not clearly identify those patients who would ultimately report regret. Still, albeit a nonsignificant difference, it should be noted that those with regret reported somewhat higher decision uncertainty at baseline, more than a 20% difference from those who did not (Table 1). In contrast to the lack of predictability of baseline factors, early evaluation of patient‐reported outcomes after just two cycles of systemic treatment strongly identified those patients who would express regret, or not, at 10–12 weeks after starting treatment. Specifically, the 3‐Item Global Index indicated the risk of regret early in the treatment course as seen in Table 1 (the 3‐IGI is derived from the eLCSS‐QL and is composed of three patient‐rated single items assessing overall quality of life, distress, and activity level).

Most treatment regimens for lung cancer are repeated every 3 weeks, as occurred in this prospective trial, and the eLCSS‐QL was assessed at the same frequency. At the 6‐week time point after starting treatment, patients had received only two cycles of treatment (generally at days 1 and 22) and had recovered from most of the immediate side effects of treatment, such as emesis or myelosuppression. Of interest, at this early point, the physicians only rated response in 7% of the patients. Those patients who experienced a 20% decline in the 3‐IGI score at 6 weeks, when compared with the baseline value, had a much poorer survival than patients not having this decline (17% vs. 53% at 1 year, p = .0025) and were much more likely to experience regret at the 3‐month mark (14% versus 1%). Treatment regret was far more frequent in those patients with the 20% 3‐IGI decline after 2 treatment cycles. The 20% decline in the 3‐IGI can be seen as a patient perceived worsening in health status. This then resulted in 17% of patients regretting their choice of treatment that had caused them harm but not their decision to pursue treatment (only 3% expressed that aspect of regret), as illustrated in the results of the different questions of the DRS given in Table 2. The 3‐IGI assesses several aspects of great importance to patients and their caregivers [18]; however, it is possible that other factors not assessed in this trial could also influence regret.

No significant differences in expressing regret were found between the enhanced care and usual care randomized groups as benefit of the decision aid for the outcomes of conflict and regret. Given the relatively small total number of patients expressing regret (15 patients, or 11.5%), it must be realized that the difference between groups would have had to have been very large to result in significant differences by randomization group. As indicated above, those patients randomized to the enhanced care group used the DecisionKEYS decision aid. In evaluating this interactive process, more than 75% of these patients and 70% of their supporters indicated that the decision aid was helpful in arriving at the choice for their cancer treatment with the doctor and that the decision aid helped them feel that they shared in the decision making for treatment.

Two strengths of the trial were that it was prospective and that it had a diverse patient inclusion. Patients participating in this study represented typical patients with advanced lung cancer in the U.S. Women composed 43% of the patients, and 47% represented minorities. Diversity in patient accrual was a goal for this study, recognizing that “racial and ethnic diversity has historically been difficult to achieve in NCI‐sponsored clinical trials,” as outlined at a National Cancer Institute 2015 workshop and in a 2014 publication [19, 20]. Including typical population groups in trials improves the likelihood that study results can be generalized to clinical practice.

A factor that contributed to the high adequacy rate (98%) of the PRO evaluation over time was the use of a simple and highly acceptable electronic computer‐assisted QL‐PRO system, the eLCSS‐QL, which had previously been tested for use in prospective trials [12]. In the initial testing of the eLCSS‐QL, which included 148 patients receiving chemotherapy, 98% of the patients found this evaluation to be acceptable, and more than 91% of the physicians indicated that this instrument could help identify earlier those patients who were benefiting or not from chemotherapy [12]. Few recent studies achieve this level of PRO adequacy over time. It is encouraging that 91% of patients who could complete the decision regret scale did so at the 10–12‐week evaluation time point.

A weakness of the trial, however, is that of the 160 patients meeting the MITT criteria, 17 patients were not evaluable for regret at the 10‐week time point in that 12 had died and 5 were too ill. These patients not evaluated for regret represent 19% of the total. It is possible that a higher percentage of this subset with such poor treatment outcomes might have expressed regret at an earlier time, which could increase the total 11.5% regret rate.

Another weakness of the study is found in assessing regret among patient supporters. First, only a minority of patients designated a supporter at the outset of the trial (43%), and of these supporters, only two thirds completed the DRS at 10–12 weeks. Although it was not a requirement for study inclusion for a patient to include a supporter, it is unfortunate that only a minority of patients included a supporter in the study. Still, this represents the largest number of supporters assessed for treatment regret in the lung cancer setting. It is interesting that the percentage of supporters expressing regret (9%) is similar to that indicated by the patients (11.5%). Given that early palliative care may contribute to patient benefit [2], it is a weakness of the trial that the percentage of patients receiving such additional care was not recorded. Moreover, a study limitation could be that it only included North American patients.

Addressing risks and benefits related to patient values and including patient‐reported factors can lead to better evaluation of patients consistent with their needs and goals. Careful review of these factors may result in better care after just two treatment cycles. The impact of interventions specifically related to decision regret should be further tested to determine if they reduce patient and supporter regret in advanced lung cancer.

Conclusions

This report documents prospectively for the first time the incidence of treatment‐related regret in patients with advanced lung cancer, and outlines that risk of regret is associated with patient‐determined worsening health status early in the course of treatment. Identifying patients at risk for regret early in treatment (before the third cycle of treatment) appears to be crucial. Counseling at that time should include a discussion of consideration of treatment change and the reason for this change.

Author Contributions

Conception/design: Patricia. J. Hollen, Richard J. Gralla, Martin Lesser

Provision of study material or patients: Richard J. Gralla, Ryan D. Gentzler, Richard D. Hall, Haiying Cheng, Balazs Halmos, Geoffrey Weiss, Jeffrey Crawford

Collection and/or assembly of data: Bethany Coyne, Jane Gildersleeve, Claudia Calderon, Ivora Hinton

Data analysis and interpretation: Martin Lesser, Jane Cerise

Manuscript writing: Patricia. J. Hollen, Richard J. Gralla

Final approval of manuscript: Patricia. J. Hollen, Richard J. Gralla, Ryan D. Gentzler, Richard D. Hall, Bethany Coyne, Haiying Cheng, Balazs Halmos, Jane Gildersleeve, Claudia Calderon, Ivora Hinton, Geoffrey Weiss, Jeffrey Crawford, Jane Cerise, Martin Lesser

Disclosures

Ryan D. Gentzler: Pfizer, AstraZeneca (C/A), Pfizer, Takeda, Jounce, Merck, Bristol‐Myers Squibb, Helsinn (RF), Targeted Oncology (H); Richard D. Hall: AstraZeneca, Takeda (C/A), Merck, Mirati Therapeutics, AstraZeneca (RF); Balazs Halmos: Astra‐Zeneca, Boehringer‐Ingelheim, Merck, Bristol‐Myers Squibb, Amgen, Genentech, Novartis, Pfizer, Spectrum, Guardant Health, Foundation One, TPT (C/A), AbbVie, Advaxis, Boehringer‐Ingelheim, Pfizer, GlaxoSmithKline, Novartis, Mirati, Merck, Bristol‐Myers Squibb, Guardant Health, Eli Lilly & Co (RF); Jeffrey Crawford: AstraZeneca, Coherus, G1 Therapeutics, Glaxo Smith Kline, Merck, Pfizer, Spectrum (C/A), AstraZeneca, Genentech, Helsinn, Pfizer (RF‐Institutional), Amgen, Enzychem (SAB), Beyond Spring, G1 Therapeutics, Merrimack, Mylan, Roche (Other‐DSMB). The other authors indicated no financial relationships.

(C/A) Consulting/advisory relationship; (RF) Research funding; (E) Employment; (ET) Expert testimony; (H) Honoraria received; (OI) Ownership interests; (IP) Intellectual property rights/inventor/patent holder; (SAB) Scientific advisory board